Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:00,540 --> 00:00:01,710
Hello, everyone.
2
00:00:01,710 --> 00:00:06,330
In this video we are diving into the exciting world of prompt engineering.
3
00:00:06,750 --> 00:00:13,560
It's the secret sauce for getting the most out of your AI art generators like Adobe, Firefly and Midjourney.
4
00:00:13,560 --> 00:00:15,390
So let's get started.
5
00:00:15,510 --> 00:00:21,690
First off, let's clarify what prompt engineering is in the context of AI art generators.
6
00:00:21,690 --> 00:00:28,710
It refers to the crafting of effective text descriptions or prompts that guide the AI in generating
7
00:00:28,710 --> 00:00:30,660
the specific images you desire.
8
00:00:31,080 --> 00:00:37,860
Prompt engineering is crucial because it allows you to control the output of the AI and get closer to
9
00:00:37,860 --> 00:00:39,960
the visual result you are aiming for.
10
00:00:39,990 --> 00:00:45,390
The better your prompts, the more accurate and satisfying your generated image will be.
11
00:00:45,600 --> 00:00:52,020
There is already a job title called Prompt Engineering in Big Companies, so let's see some tips and
12
00:00:52,020 --> 00:00:53,700
tricks for your prompts.
13
00:00:54,150 --> 00:00:56,910
Tip number one be specific.
14
00:00:57,060 --> 00:01:04,720
Instead of writing a cat, you might say a fluffy orange tabby cat sitting on a windowsill at sunset.
15
00:01:04,870 --> 00:01:09,970
The more details you provide, the better the AI can visualize what you are looking for.
16
00:01:10,240 --> 00:01:13,300
My second tip is to try different phrasings.
17
00:01:13,330 --> 00:01:19,540
If the first prompt doesn't give you the result you want, try expressing the same idea in a different
18
00:01:19,540 --> 00:01:20,080
way.
19
00:01:20,110 --> 00:01:21,900
Use different keywords.
20
00:01:21,910 --> 00:01:29,050
For instance, instead of a medieval castle at night you could try a stone fortress under the moonlight.
21
00:01:29,470 --> 00:01:33,340
Tip number three Experiment with abstract concepts.
22
00:01:33,580 --> 00:01:37,600
AI generators can handle more than just literal descriptions.
23
00:01:37,600 --> 00:01:43,900
You could try something like the feeling of nostalgia or the concept of infinity.
24
00:01:44,740 --> 00:01:48,700
Fourth tip use adjectives and adverbs wisely.
25
00:01:48,910 --> 00:01:55,510
Words like gigantic, sparkling, peacefully can significantly influence the AI's output.
26
00:01:56,200 --> 00:01:57,200
Final tip.
27
00:01:57,220 --> 00:01:59,260
You might already be asking for.
28
00:01:59,260 --> 00:02:03,700
How can I remember all of these adverbs and fancy words before creating something?
29
00:02:03,730 --> 00:02:08,640
Well, we are living in 2023, so there is an AI tool for that.
30
00:02:08,650 --> 00:02:13,480
You can use a language model like ChatGPT to create prompts.
31
00:02:13,870 --> 00:02:15,700
Let's see how we do that.
32
00:02:15,730 --> 00:02:22,330
I will go to ChatGPT and say, Write me a generative art prompt for futuristic photography.
33
00:02:25,800 --> 00:02:28,410
Oops, that's not what we expected, actually.
34
00:02:28,440 --> 00:02:31,020
We are not prompting Firefly in this way.
35
00:02:31,020 --> 00:02:31,740
Right.
36
00:02:31,860 --> 00:02:32,250
Okay.
37
00:02:32,250 --> 00:02:39,360
So ChatGPT apparently doesn't know how Adobe Firefly expects the prompts, but we can train it.
38
00:02:39,390 --> 00:02:45,540
We can provide the formula of a prompt to ChatGPT so that it knows how it should write.
39
00:02:45,570 --> 00:02:48,360
Let me show you one of my chats with ChatGPT.
40
00:02:48,750 --> 00:02:56,580
I first define the formula for this formula subject of the photo three Descriptive keywords of the image
41
00:02:56,580 --> 00:03:02,100
camera type camera lens type time of the day style of the photograph.
42
00:03:02,100 --> 00:03:08,550
It understands the structure and already gave me a prompt, which is a Golden Gate bridge.
43
00:03:08,550 --> 00:03:13,230
And these are the descriptive words, the camera type and also the lens type.
44
00:03:13,260 --> 00:03:19,710
Time of the day is sunset in this example and style of the photograph is long exposure.
45
00:03:19,890 --> 00:03:21,510
So this is perfect.
46
00:03:21,510 --> 00:03:23,610
Let's try this with Adobe Firefly.
47
00:03:23,850 --> 00:03:27,820
I will just copy and paste the prompt here and wait a bit.
48
00:03:28,860 --> 00:03:29,910
And there you go.
49
00:03:31,100 --> 00:03:32,300
Okay, this is perfect.
50
00:03:32,300 --> 00:03:33,410
But there is more.
51
00:03:34,040 --> 00:03:41,480
Then I say the name of this formula is F1 stands for Adobe Firefly, Formula one and type.
52
00:03:41,480 --> 00:03:48,260
Yes, if you understand it says it understands the formula and I will provide a new one with the name
53
00:03:48,410 --> 00:03:49,340
F2.
54
00:03:50,220 --> 00:03:59,250
This time we can include adjective subject, doing an action lighting, and you can also add some static
55
00:03:59,250 --> 00:04:04,470
words like detailed and realistic if you want to include in all of the prompts.
56
00:04:05,510 --> 00:04:11,000
And you can say in the style of a movie or artist at the end of the prompt.
57
00:04:11,420 --> 00:04:14,930
Okay, Now it says, I understand the second formula as well.
58
00:04:14,930 --> 00:04:23,420
And now you can say, Give me a lion image using AF two and look at this prompt an image of fierce lion
59
00:04:23,420 --> 00:04:26,090
roaring under the dramatic twilight.
60
00:04:26,120 --> 00:04:31,070
Detailed, realistic, just like we said in the style of The Lion King.
61
00:04:31,100 --> 00:04:32,600
And I like that part.
62
00:04:32,600 --> 00:04:38,390
And Salvador Dali, I am not sure Firefly is artist aware at the moment.
63
00:04:38,420 --> 00:04:42,410
Artists Styles works better with Midjourney, but we can try.
64
00:04:42,440 --> 00:04:46,850
Let's try with Adobe Firefly and these are the image we get.
65
00:04:47,450 --> 00:04:50,390
By the way, you don't need to ask one prompt at a time.
66
00:04:50,390 --> 00:04:55,460
For example, you can say Write me five prompts for lion image using a fun.
67
00:04:55,970 --> 00:05:02,960
Okay, so this is the concept of defining formulas and using language models to create your prompts.
68
00:05:03,200 --> 00:05:08,460
So I strongly recommend you to define your own formulas according to your needs.
69
00:05:08,460 --> 00:05:14,130
For instance, you can define one for realistic photographs and one for digital artworks.
70
00:05:14,130 --> 00:05:16,410
And that's totally up to you.
71
00:05:16,440 --> 00:05:19,050
All right, guys, that was the Tips and Tricks video.
72
00:05:19,050 --> 00:05:21,360
I hope it was helpful and valuable.
7109
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.